Neuroimaging for Migraine: The American Headache Society Systematic Review and Evidence-Based Guideline

Randolph W. Evans, MD; Rebecca C. Burch, MD; Benjamin M. Frishberg, MD; Michael J. Marmura, MD; Laszlo L. Mechtler, MD; Stephen D. Silberstein, MD; Dana P. Turner, MSPH, PhD


Headache. 2020;60(2):318-336. 

In This Article


The initial search yielded 2269 publications. After review of the titles and abstracts, 85 articles were selected by RE and BF for full screening, and 25 meeting the inclusion criteria were chosen for final review and extraction. After excluding 4 articles that were determined to be ineligible after a full review, 21 articles were retained for inclusion in the synthesis of results. Studies of cluster headache were also excluded after determining that none met the review inclusion criteria. The updated search in 2018 yielded 2 new articles that met the inclusion criteria. A total of 23 articles were therefore included in the synthesis (See Table 1). The majority of studies were retrospective cohort or cross-sectional studies. There were 4 prospective observational studies.

While most of the articles (18/23) focused on the use of neuroimaging in the setting of a normal neurological exam and diagnosis of migraine, 5 studies specifically focused on migraine subjects with more worrisome signs or symptoms raising concern for a secondary cause. These included 2 studies of subjects with migraine presenting with acute changes in headache pattern or focal neurological abnormalities,[17,18] long-standing serious migraine (average 20 years) or with "permanent neurological sequelae,"[19] patients with a diagnosis of "non-epileptiform basilar artery migraine,"[20] or disabling migraine requiring hospital admission including hemiplegic or vertebrobasilar migraine.[21] A few studies preceded the recognition of important secondary headaches such as reversible cerebral vasoconstriction syndrome, and the link between migraine and white matter lesions.[20,22] Outpatients seeking care for headache in the clinic were the most common subjects. A few studies included both typical migraine subjects, with and without aura, as well as those with complicated or hemiplegic migraine.

CT Neuroimaging Studies

Ten studies reviewed the results of CT scans in subjects with migraine (Table 2). In some cases, the authors focused on all abnormalities, while others reported only clinically meaningful findings. In 2 studies, Cala and Mastaglia reported 2 subjects with masses felt to be glioma, 6 with cerebral infarcts and 6 with periventricular edema among their 94 subjects presenting with acute changes in headache pattern or the presence of focal neurologic abnormalities.[17,18] Hungerford et al found abnormalities in almost half of their subjects (25/53) including 6 with generalized atrophy, 8 with focal atrophy, and 5 with other abnormalities.[19] Cuetter and Aita reported only 1 clinically significant abnormality in their 435 subjects with migraine: a choroid plexus papilloma of the 4th ventricle.[21] Mathew et al reviewed results in hospitalized subjects with significant complications and found abnormalities in 10/29 subjects including 4 with ventricular enlargement, and 6 with 1 or more areas of low density.[23] Akpek retrospectively evaluated 592 subjects who had CT for migraine and found that 40 (8%) had minor abnormalities but none had major abnormalities. The abnormalities included 16 with cerebral or cerebellar atrophy, 12 with chronic ischemia, 9 with possible pseudotumor cerebri, 3 with venous angioma, 2 with empty sella, 2 with ventricular asymmetry, 1 with basal ganglia calcifications, and 1 with a subcutaneous fibroma.[24] Kahn et al[25] reviewed findings in 1111 patients presenting at 2 large teaching hospitals for acute non-traumatic headache; About 10.8% was abnormal. Abnormalities included acute infarction (44, 4.0%), primary neoplasm 18, (1.6%), subarachnoid hemorrhage 14 (1.3%), subdural hematoma 12 (1.1%), metastatic neoplasm 12 (1.1%), vascular abnormalities such as aneurysm or AVM 11 (1.0%), and hydrocephalus 9 (0.8%). Sargent et al[26] studied 82 subjects migraine and found 20 of 82 had abnormalities such as generalized or focal atrophy, with increasing prevalence in older subjects. Masland et al studied 136 subjects with migraine and during acute migraine attacks[27] and reported that 19 (14%) had abnormalities including 14 with cerebral atrophy, 2 with evidence of small infarcts, 1 with AVM, 1 with astrocytoma, 1 with meningioma, and 1 had a saccular aneurysm of the anterior communicating artery. CT scans did not show abnormalities during migraine attacks, and the 2 patients with severe intracranial pathology also had recent clinical changes including progressive hemiparesis and focal seizure suggesting a need for neuroimaging. Valenca et al reviewed findings in 78 patients: 34 with migraine, 35 with tension-type, and 9 with both disorders.[28] Abnormalities were common but mostly incidental including inflammatory sinus disease (19.2%), cysticercosis (3.9%), unruptured cerebral aneurysm (2.6%), basilar impression (2.6%), intracranial lipoma (2.6%), arachnoid cyst (2.6%), empty sella (2.6%), intracranial neoplasm (2.6%), and colloid cyst (1.3%).

Many of these early CT studies report frequent abnormalities, findings such atrophy, chronic ischemia, basal ganglia calcifications, sinus disease or ventricular enlargement, very few of these changed the diagnosis or led to intervention. Many of the reported abnormalities such as atrophy likely reflect normal aging.[29] The prevalence of unruptured intracranial aneurysms in these studies was similar to the general population.[30] While there were some serious abnormalities, in one study, the 2 patients with severe pathology had a progressive hemiparesis and another with seizure.[27] Other abnormalities occurred among subjects in studies specifically focused on hospitalized subjects [25] or in those with new changes in headache or exam abnormalities.[18]

MRI Neuroimaging Studies

Nine studies reviewed the results of MRI studies in subjects with migraine (Table 3). Cooney et al retrospectively analyzed 185 consecutive patients with migraine to correlate MRI abnormalities such as white matter lesions with patient demographics and clinical features.[31] Only 30 of the 185 (16%) had white matter abnormalities, with a higher prevalence in subjects over 50 and those with risk factors such as hypertension, heart disease, or diabetes mellitus.[31] Gozke et al[32] studied 45 patients with migraine: 20 with aura and 25 without. White matter foci were noted in 13/45 subjects (28.8%) and significantly more common in those with aura (8/20, 40%) compared to those without (5/25, 20%). There was 1 patient each with frontoparietal cortical atrophy and heterotopy – felt to be incidental findings. Honnigsvag reviewed MRI findings in a population-based cross-sectional study of adults aged 50–65 who had participated in previous Nord-Trøndelag Health Studies (HUNT).[33] Patients with any headache disorder had a higher rate of any intracranial abnormality as compared with the non-headache population (29% vs 22%), including major (11% vs 10%) and minor (17% vs 13%) categories. However, when white matter hyperintensities were removed from the analysis, this association disappeared. While abnormalities were common in both groups including venous angioma, multiple sclerosis, carotid disease, AVM, and pituitary tumor, there was no significant difference between headache and headache-free groups.[33] Osborn et al[34] reviewed MRI findings in a relatively younger group of migraine subjects, with a mean age of 29.8 years. They detected white matter lesions in 5 of the 41 patients (12%) which were less common in those under 40 years old (5.5%). Prager and Rosenblum retrospectively reviewed 77 subjects with migraine, to determine if white matter abnormalities were associated with clinical features such as diagnosis, sex, age, number of years with symptoms, and history of vasoactive medication.[35] These abnormalities were common in subjects with (44%) and without (47%) aura, and more common with advancing age. Soges et al[36] investigated 24 patients with migraine, including 7 without aura and 17 with aura. They determined that white matter lesions were common in those with aura (4/7, 57%) and those without aura (7/17, 41%). They also reported addition 3 large cortical abnormalities in the group with aura multiple bilateral focal white matter lesions in another subject. Wang et al[37] retrospectively reviewed neuroimaging findings in 402 subjects referred for headache by a neurologist including 161 subjects with migraine. Major abnormalities were significantly less common in those with migraine, the only abnormality was a petrous apex cholesterol cyst in a 58-year-old woman. Mullally and Hall[38] sought to determine if patients who request neuroimaging for headache would be more likely to have serious abnormalities. Of 100 subjects with a migraine diagnosis, including 41 with chronic migraine, and normal neurological examination, most scans were normal (82%) but 17% had insignificant abnormalities. One patient had a serious abnormality: a meningioma that eventually required surgery and radiation therapy. Jacome and Leborgne[39] specifically studied 18 patients with a diagnosis of "non-epileptiform basilar migraine" and found an abnormality in 12/18. Of these, 6 had mildly enlarged sulci, 1 had moderately large ventricles and T2 focal signal abnormalities, 1 had cerebellar vermal hypoplasia, 1 had basal ganglia calcifications, and 2 had a single white matter lesion.

In addition to abnormalities noted on CT, MRI studies frequently showed white hyperintensities and occasionally pituitary abnormalities. These studies primarily focused on outpatients with migraine with no indication of exam abnormalities. Other than the patient with meningioma, there were no serious abnormalities requiring intervention.

Studies Using Both CT and MRI

Four studies included subjects with either CT or MRI imaging. Wang[40] compared neuroimaging results between 1070 subjects with headache (including 665 with migraine) and 1070 healthy gender and age matched controls, without "red flags" or abnormal exam findings. None of the 382 subjects undergoing CT had significant abnormalities. Of the 688 subjects receiving MRI, only 4 subjects with headache had significant imaging abnormalities (0.58%), including 3 of the migraine subjects (0.67%). Abnormalities included 2 subjects with hydrocephalus and 2 with tumors of the throat and nose. There were 5 abnormalities in the healthy controls (0.73%) which was not significantly different.[40] Cull[41] studied CT and MRI abnormalities in 69 subjects presenting with "late-onset migraine" starting after the age of 40 with the majority of subjects (86%) having aura. When available, carotid ultrasound and laboratory testing was also reviewed. About 93% had normal neuroimaging. Abnormalities included 4 subjects with evidence of a previous cerebral infarction, and 3 with mild-moderate carotid atheroma on ultrasound scanning. Kuhn and Shekar reviewed CT and MRI findings in 74 pediatric and adult subjects (age 9–39, mean 28) with classic migraine. MRI revealed multiple foci of bright signal on T2 MRI in 19 of the 74 subjects (26%) which were not detected on CT. Focal or generalized ventricular enlargement or sulcal prominence was present in 26 subjects both on CT and MRI. One patient with homonymous hemianopsia had an occipital lobe infarct – seen on both MRI and CT.[42] Clarke et al reviewed CT and MRI findings in sequential new patients who had neuroimaging over a 5-year period.[43] Of the 167 patients with a diagnosis of migraine, only 2 (1.2%) had significant abnormalities and both had known possible secondary causes of headache. One abnormality was a Chiari malformation in a patient with a history of Moya-Moya disease and extracranial-intracranial bypass surgery, and another presenting with blurred vision and early morning headache with known Dandy-Walker syndrome was found to have a blocked shunt (Table 4).

These studies with CT or MRI demonstrated frequent abnormalities but with similar incidence to those in the general population or control groups. Exam abnormalities such as vision loss or history of neurosurgery predicted significant pathology in CT or MRI.

Quality Assessment

The quality of studies included in this review was generally poor. Scores on the Newcastle-Ottawa Scale for cross-sectional studies ranged from 0 to 6 (Table 5). The majority of studies received either a 0 (9 studies) or a 1 (6 studies). Only 1 study each received a 5 or 6. No studies received the score of >7 required to be considered high quality. Only 1 study, Honningsvag et al,[33] was rated as being truly representative of the population of interest. The majority of studies included in this review were based on either consecutive recruitment from a healthcare setting, or were convenience samples. No paper described a sample size calculation. Three studies[20,22,33] reported that outcomes were determined by independent blind assessment. Most radiographic outcomes were determined by radiologists who were aware of the indication for imaging. Based on the methodological flaws of the included studies, further good quality studies are needed.